|By Roger Strukhoff||
|February 10, 2014 09:43 AM EST||
I want to build a cloud-driven datacenter. I want to build it alongside a newly installed fiber-optic network near my office on an old college campus in Northern Illinois.
I'm advising a company that has ambitious plans to reach a large number of consumers by end of 2015 with a social-media service that appears to me to be unique and wonderful. They/we are going to need the power of several hundred servers, maybe more.
As a long-time writer about cloud computing and one of the guys behind Cloud Expo, my natural inclination is to buy more public cloud. We're already running instances. As Conference Chair of the WebRTC Summit and Things Expo, I had better be a leading-edge sort of person.
But why not build? Virtualize our new servers and attendant resources, measure the services, and back them up with more public instances. Create a local, cloud-redundant, expandable hybrid cloud.
I've run the numbers, and they tilt in favor of building, over time. Plus, with this option we can employ some people locally (building and operating), become a major customer of the local utility co-op, and be responsible for our own outages.
Make It Someone Else's Problem
Or maybe I should just punt. Let this problem remain with the firm's current CTO, who's based in Chicago. He'll probably just buy more cloud until he realizes his opex budget becomes, in fact, more onerous than the combination of depreciated capital investment and operating costs we'll have in our hybrid cloud here in the boondocks.
If I build, my life will become an endless nightmare of details, decisions, screw-ups, and complexity. But if I buy, things won't be so different. It's a complex challenge scaling up to a couple of magnitudes beyond current needs.
Only if I punt will my life maintain the "stability" it currently has.
PaaS, IaaS, & IoT
Oh, we need to include a serious, new software-development effort as we scale up. We need PaaS. Or, in today's evolving terminology, has PaaS faded into the world of IaaS? Our programmers won't be here in the boondocks. Will they be better off with a public-cloud resource, or this local datacenter that I want to build?
We anticipate most people accessing our site with mobile devices, both from their homes and when they visit facilities we'll be running. We've thought about grabbing their MAC addresses so we can watch traffic patterns. Whether we do this or not, I can see an Internet of Things in our future as we watch the ebb and flow of many, many people.
Back to basics. Back to RAS. And back to the idea that I really want to build this thing. So how about a pilot project? I think I'll get that first rack, see if I can locate and run it properly, then build from there.
It should be fun.